The outcome

Publish a clear synthetic performance without misrepresenting a real person or losing control of their likeness and voice.

Step by step

A workflow you can repeat.

  1. 01

    Document the speaker or rights holder's consent, allowed script, channels, duration, audience, territories, revocation path, and synthetic-media disclosure.

  2. 02

    Prepare a front-facing authorized image and clean authorized audio, removing metadata and private background details that are not needed.

  3. 03

    Generate a short low-resolution test with the chosen avatar model, aspect ratio, neutral motion prompt, and no public sharing.

  4. 04

    Review pronunciation, lip-sync, identity, gaze, teeth, hands, gesture, emotion, captions, background, and whether the performance implies unapproved claims.

  5. 05

    Correct the source or prompt, render only the approved take, add visible disclosure and accessibility, obtain final sign-off, and delete unused identity assets where appropriate.

Working standard

What good use looks like.

  • Get specific written consent.
  • Test with a short clip first.
  • Keep disclosure attached to the media.

Official references

Check the current product documentation.